ABSTRACT Repeated major depressive episodes are particularly problematic for older adults who have a more brittle recovery than younger adults. Our data show that, despite antidepressant treatment, almost 60% of remitted older adults experience recurrence within four years. Beyond simply relying on past history and reported current stress, it is unclear what neurobiological factors are prospectively associated with recurrence risk, when these factors trigger recurrence, and how they contribute to the high rates of cognitive impairment observed in late-life depression (LLD). Using a model of network homeostasis, we posit that depressive episodes are characterized by disrupted homeostasis in key neural networks involved in affect regulation and cognitive function. Our preliminary data indicate that treatment non-remitters have residual functional network alterations and high network instability (higher fluctuations in temporal signal-to-noise ratio). We hypothesize that remitters with residual functional network alterations and greater instability remain at high risk of recurrence with subsequent stress exposure. This disequilibrium contributes to subsyndromal symptoms followed by full recurrence. These processes may also contribute to the higher rate of cognitive impairment and decline observed in LLD. Our groups have reported elevated rates of cognitive decline in remitted LLD and an association of recurrence with accelerated brain aging. We hypothesize that greater neural reactivity to stress may accelerate brain aging and cognitive decline and that deficits/variability in performance on tasks dependent on ECN may serve as markers of network alterations and signal increased recurrence risk. The goals of this study are to A) identify neurobiological factors that predict recurrence risk, and B) examine how cognitive performance changes are both influenced by these same neurobiological factors and also predict recurrence risk. Our approach is to conduct a three-site, two-year longitudinal study of remitted LLD and never-depressed elders. Every 8 months we will conduct laboratory assessments, including clinical, cognitive and neuroimaging assessments and an in-scanner stress paradigm, along with burst ecological momentary assessments (EMA) of mood variability, stress exposure, cognitive performance, and passive actigraphy. As an exploratory goal, we will examine whether continuous ecological monitoring of mood and activity can provide early detection of recurrence. A subgroup will be continuously monitored by EMA and actigraphy for state shifts (persistent worsening) or variance shifts (increased variability) in symptom severity. When shifts in mood symptoms are identified, they will engage in ad-hoc clinical and neuroimaging testing. Results from this study may be translated in clinical practice through the future development of easy-to-use platforms (e.g. apps) that signal to clinicians increased risk of impending recurrence, thus allowing for swift therapeutic intervention.